Multi-dimensional representation learning for word retrieval – We develop a new, yet complementary method of learning an attention-based topic model (DASMM) for multi-dimensional representations. Traditionally, the task of representing a new topic is treated as a single- or multiple-dimensional representation task. We solve the multi-dimensional representation learning problem in an alternating direction method of learning the topic labels, and show that even a single word can significantly help in learning a multi-dimensional representation. We prove that the learning problem can be solved efficiently via the non-convex convex relaxation of the convex relaxation equation over multi-dimensional representations. In a large amount of experiments, we evaluate the effectiveness of our method in various domains and show the benefits of this method.

I present the model and algorithm for the task of automatic and collaborative semantic segmentation of a patient’s hands (with the help of a digital medical record). It is a challenging task due to large variations of the hands and pose of patients, such as people wearing different uniforms. While hand pose reconstruction has become the focus of recent research, the most recent research focuses on fine-grained segmentation of hand gestures by using a single image. In this paper, we propose a novel deep learning based method to segment the hand poses using a neural machine translation (NMT). We present a deep neural language model to directly improve the hand pose reconstruction. The proposed method employs a deep convolutional neural network (CNN) to classify hand poses. This model is trained with hand pose prediction and segmentation tasks as pre-processing steps. We demonstrate that the proposed method outperforms state-of-the-art hand pose reconstruction approaches on a variety of hand pose baselines by over 40% accuracy on all tasks tested.

Deterministic Kriging based Nonlinear Modeling with Gaussian Processes

Distributed Sparse Signal Recovery

Multi-dimensional representation learning for word retrieval

Probabilistic Modeling of Time-Series for Spatio-Temporal Data with a Bayesian Network Adversary

Deconvolutional Retinex and Neural Machine TranslationI present the model and algorithm for the task of automatic and collaborative semantic segmentation of a patient’s hands (with the help of a digital medical record). It is a challenging task due to large variations of the hands and pose of patients, such as people wearing different uniforms. While hand pose reconstruction has become the focus of recent research, the most recent research focuses on fine-grained segmentation of hand gestures by using a single image. In this paper, we propose a novel deep learning based method to segment the hand poses using a neural machine translation (NMT). We present a deep neural language model to directly improve the hand pose reconstruction. The proposed method employs a deep convolutional neural network (CNN) to classify hand poses. This model is trained with hand pose prediction and segmentation tasks as pre-processing steps. We demonstrate that the proposed method outperforms state-of-the-art hand pose reconstruction approaches on a variety of hand pose baselines by over 40% accuracy on all tasks tested.